II
DSS COMPONENTS
Decision Support Systems for Business Intelligence
by Vicki L. Sauter
Copyright © 2010 John Wiley & Sons, Inc.
DATA COMPONENT
Business analytics, and thus business intelligence efforts, are dependent upon data. If there
are no data, there are no business analytics. If there are no business analytics, then we
cannot exploit the edge of understanding the business, its performance, and its context,
which in turn means we cannot improve our decision making. All of that suggests that the
performance of our corporation will not be up to its potential. In fact, in today's competitive
world, it may mean that the organization may no longer exist.
Hence, before we can talk about how to make models more understandable or how to
project the appropriate information to the screen, it is critical to discuss how to know what
data need to be included in the DSS. Before we can do that, we need to define
data
and its
associate,
information.
Data
are
things known or
assumed.
The term generally refers to facts and/or figures from
which conclusions can be drawn. For example, the raw counts of walnut consumption and
cholesterol levels discussed in Chapter 2 represent data. Similarly, the cost of commercial
time and the distribution of viewing audiences of television programs represent data
to
those
making marketing plan choices. Details about shipping procedures, cost, and reliability of
various haulers represent data relevant to the development of a logistics plan.
However, these are not the only kinds of details that might be considered data for the
purposes of
DSS.
When making choices, some decision makers value the opinions of trusted
colleagues. For example, when purchasing managers consider new, unknown vendors, they
often seek opinions regarding service and reliability from colleagues at other corporations
who have purchased from those vendors. They would not use these opinions solely but
would use them to enrich a cost model developed from more objective data. Similarly,
Decision Support Systems for Business Intelligence
by Vicki L. Sauter
Copyright © 2010 John Wiley & Sons, Inc.
DATA COMPONENT
when developing a long-range plan, a CEO enlists knowledgeable subordinates to gauge
the expected changes in regulations, governments, vendors, competitors, and clients over
a 20-year period. These opinions are melded with quantitative models, which alone do not
provide reliable long-range forecasts, as the basis of
a
long-range estimate of the company's
needs.
In each of these cases, opinions and judgments are used as inputs to a choice process.
They supplement standard "objective" data to represent aspects of the choice that would
otherwise be lacking. Since the DSS is intended to
support
the choice process, it must
accommodate such subjective data and opinions and provide efficient ways of searching
for and using these data.
For other decisions, decision makers might need data that are not stored in conventional
ways.
For example, decision makers considering the choice of textiles for the manufacture
of furniture believe the support provided by pictures is superior to that provided by verbal
descriptions of the colors, patterns, and textures. Images supplement data such as price,
vendor, or shrinkage that would be accessed in a standard fashion. Decision makers con-
sidering a large-scale disaster relief plan might need a video of the affected area to assess
the problems and needs of an area fully. Such a video needs supplementary geographical
information systems support to assess land use, damage estimates, and population statistics
for each affected area. Or, a symphony music director might find it beneficial to have
audio files of possible selections to help select a balanced and appealing program. With
the audio data, the music director might combine data, including programs in which the
piece has been used, audience size, reviews, and comments, to develop models that max-
imize the number of new compositions played by an orchestra while still being sensitive
to the expected composition of the audience, thus pairing new selections and established
favorites in a pleasing fashion.
With virtual-reality technology, decision makers might also access "experiences" be-
fore they select alternatives. For example, city planners might make use of virtual reality
in positioning new buildings or green spaces, including the evaluation of the aesthetics and
access. Similarly, fashion collections can be modeled using virtual reality (replicating the
variety of poses and settings that might happen at actual fashion shows) in order to get a fast
opinion of designers and/or customers prior to their announcement. Or, a logistics planner
could use virtual reality to evaluate space needs, safety issues, or production principles.
One of the purposes of the DSS is to transform these data into information that can
help the decision maker. While data represent things known or assumed, information refers
to processed data or "acquired knowledge." Processing can be a summarization (either
numerical or graphical) or the output from one or more models. For example, scores on
an exam in a particular class represent data; each score represents performance by the
corresponding individual. However, they do not represent information. This list does not
help you, as an individual student, decide how to respond to your performance on the exam.
Once the data are processed, however, they do support your decision. With a computation
of class mean and standard deviation or the identification of cutoffs associated with each
letter grade, students can decide whether they performed at a personally acceptable level,
whether they should study harder, and whether they should drop the class.
In the simplest terms, if
the
data are not in and of themselves information, or if
the
data
cannot be transformed into information, then they should not
be
included in the database. As
you can imagine, this leaves a great deal of ambiguous latitude. Returning to basics reminds
us that the goal of business intelligence is to study historical patterns and performance so as
to predict the future and improve the organization's response to future events. That means
that the data need to represent practical indicators of what is happening in the organization,
DATA COMPONENT
indicators of when changes occur, and indicators of when and how actions need to be taken.
The data need to reflect historical, current and predictive views of the organization and its
environment.
There are three approaches to operationalization of the description. The first is to
take a normative approach to the information needs: What information
should
the de-
cision maker want to make this type of decision? This assumes that which meets the
standard guidelines for making a particular decision will be useful in a given decision-
making situation. It is the material taught in business administration courses, advo-
cated in textbooks, or specified in company or professional guidelines or standards.
For example, when making a decision regarding inventory policy, standard operations
management texts advocate knowing the distribution of demand for some time period,
the expected demand for that time period, the costs of ordering the product, and the costs
of holding the product in inventory. Hence, the normative approach says that those are the
kinds of information that should be included in an inventory support system.
Few decision makers approach choices as straightforwardly as is taught in business
courses, and so the normative approach alone is not sufficient to guide the database devel-
opment. Most decision makers believe the theoretical approach to solving their problems
is not sufficient to respond to the variety of issues encountered in real decision contexts.
Specifically, these approaches do not address the question of how to make a decision if the
data are not available or are not sufficient or how to include necessary political factors in
the process.
So,
the designer of
a
DSS must also use a subjective approach to judging the usefulness
of information. Here subjective refers to the perspective of
the
decision makers—what they
think
will be useful. This allows decision makers to specify the full range of information
they might consider in the process, whether or not it is specified by the normative approach.
For example, decision makers might indicate that when deciding how much of a product
to order for inventory, they must address a wide range of issues in addition to cost. For
example, the decision of how much of an item to acquire might mean making trade-offs
between this order and the availability of other products (because of competition for space
or capital) or opportunity costs. Further, the question of how many items to have on hand
might be linked to image considerations. This would tell the designer of a DSS to include
these additional factors in the database for the DSS.
A third viewpoint is the realistic approach, which asks whether decision makers will
use particular information if it is included in the database. Some decision makers might
not have confidence in sophisticated models, either because they do not understand or
appreciate them, because they have had bad experiences with them in the past, or because
it might be politically difficult to use them in certain contexts. Designers of DSS should
be realistic about whether such information will therefore ever be used. If it is not likely
that decision makers will use it, then designers need to evaluate how much including the
information will cost and whether that money, time, or opportunity might be put to better
uses.
The DSS designers realize that choices regarding inclusion of data in a DSS involves
compromise between the normative view of decision making, the subjective view of what
is useful, and the realistic view of whether and how information can really be used in the
choice process. Sometimes this means that data are dropped from the system while other
times it means that parallel data (more palatable to the decision makers) are included in the
system. Still other times, compromise means adding help screens and warning messages to
make it easier for decision makers to use the information.
DATA COMPONENT
SPECIFIC VIEW TOWARD INCLUDED DATA
So,
what needs to be included? Most DSS first and foremost include financial information.
These reflect quantitative data indicating costs and revenues by organizational units or prod-
ucts or
regions.
Such data allow
a
manager
to
evaluate returns on investment and profitability
indices. These and other financial analyses do provide some insights into the business and
often are the dominant measure of performance. Most markets place emphasis on revenues,
net profit, and earnings per share. In addition, these financial measures are consistent across
an organization, even one that is highly diversified in products or operations.
However, financial data only provide one part of the picture. First, the financial data
might not reflect all that is of value within an organization. Even when they can reflect the
value, since they are outcome values, they tend to be lagged with regard to the activity that
caused them. If an organization is going to use analytics effectively to manage the business,
they need to understand the drivers of the activities that can be manipulated to improve the
ultimate financial outcomes. Relevant information also reflects the operational perspective,
the technical perspective, the schedule perspective, the legal and or ethical perspective, and
the political perspective of the choices that are being considered. Clearly, this requires a
wide range of information. So, what information does one select?
Gartner, in a report in 2006, identified a value model to help designers of DSS to
know what information to include. (Smith, Apfel, and Mitchell 2006). A similar matrix
applicable to a university is shown in Figure
3.1.
These measures focus on the controllable
activities within
the
demand management, supply management, and support services aspects
of the corporation. The Gartner research reports are a source for specific measures and
methodologies for measurement. While this is a nice starting point from which to get some
ideas about what to measure, even the authors indicate that it must be supplemented with
company-specific measures of what is important.
To measure important factors, managers need key performance indicators (or KPIs)
which reflect how closely the organization is moving toward its strategic direction. For
example, if the strategy of the organization is to increase the number of customers, three
Business
Aspect
Demand
Management
Supply
Management
Support
Services
Aggregates
Market
Responsiveness
Recruitment
Responsiveness
Faculty
Classes
Development
Measures
Number of
Students
Attracted
Forecast
Accuracy
% coverage
full time
Course
Evaluation
Measures
Infusion of
New
Methods
Quality of
Students
Attracted
Cost to
Recruit
Students
% coverage
academically
qualified
Fill Rate
New Classes
Undergraduate
-Graduate
Ratio
Student
Retention Rate
Turnover
Assessment
Results
Technology
Support
Channel
Success
Rates
Grad Rate
Ratio of
Senior-Junior
Faculty
Learning
Goals Ratio
Library
Support
Figure 3.1. Values Matrix.
